On 9/24/20 10:03 AM, Richard Biener wrote:
The symbols are added to offload_vars + offload_funcs.
In lto-cgraph.c's output_offload_tables there is the last chance
to remove now unused nodes ? as once the tables are streamed
for device usage, they cannot be changed. Hence, there one
has
node->force_output = 1;
[Unrelated: this prevents later optimizations, which still
could be done; cf. PR95622]
The table itself is written in omp-offload.c's omp_finish_file.
But this is called at LTRANS time only, in particular we seem
to stream the offload_funcs/vars array, marking streamed nodes
as force_output but we do not make the offload table visible
to the partitioner. But force_output should make the
nodes not renamed. But then output_offload_tables is called at
the very end and we likely do not stream the altered
force_output state.
So - can you try, in prune_offload_funcs, in addition to
setting DECL_PRESERVE_P, mark the cgraph node ->force_output
so this happens early? I guess the same is needed for
variables (there's no prune_offloar_vars ...).
As it accesses global variables, I could do just the same
with the variables – but it did not seems to have an effect.
Following Jakub's suggestion, I also added
__attribute__((used))
to the tree belonging to both tables in omp-offload.c's omp_finish
but that did not help, either.
I think both the 'used' and 'force_output' are red herrings:
after all, the tables and the referrenced funcs/vars are output;
the problem is 'just' that they end up in different ltrans
while not being public. – Thus, some property ia wrong
during building the cgraph or when it is partitioned into ltrans.
Any additional suggestion to try?
Tobias
-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstraße 201, 80634 München / Germany
Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Alexander
Walter
